Support keyword-only arguments (rebased)

[skoslowski]

This is a duplicate of #1196 rebased to current master resolving the merge conflicts on the existing PR.

(Fixes #1194)

This adds support for a py::args_kw_only() annotation that can be
specified between py::arg annotations to indicate that any following
arguments are keyword-only. This allows you to write:

m.def("f", [](int a, int b) { /* ... */ },
      py::arg("a"), py::args_kw_only(), py::arg("b"));

and have it work like Python 3's:

def f(a, *, b):
    # ...

with respect to how a and b arguments are accepted (that is, a can
be positional or by keyword; b can only be specified by keyword).

[TheButlah]

+1 for this feature

[cpuhrsch]

I'd love this as well!

[cpuhrsch]

Ping: Are you planning on merging this soon? :)

[cpuhrsch]

@gst - are there any updates to this?

[cpuhrsch]

Pinging the creator of this PR @skoslowski

[skoslowski]

@cpuhrsch, I am not a maintainer of pybind11, so there is nothing I can do. This PR only exists to help get the original code merged.

I have been wanting this for a while now and it has dragged on for too long. So, for now, I have fallen back to using a private fork with this patch included.

[cpuhrsch]

Mentioning @wjakob since this PR was approved a while ago and seems very useful in allowing pybind11 to further cover the python function-calling paradigm.

[izdeby]

This would be a great addition to pybind. Is there anyway this can be merged, @gst?

[gst]

I dont have merge privilege here otherwise I'd have already merged ;)

@wjakob is the last person having done a commit 10 days ago.. ping ?

[jagerman]

I like this addition, but some minor bikeshedding: Thoughts on changing the tag to something shorter, perhaps py::kwonly{}? The name feels a little verbose to me currently:

m.def("f", [](int a, int b) { /* ... */ },
        py::arg("a"), py::args_kw_only(), py::arg("b"));

vs:

m.def("f", [](int a, int b) { /* ... */ },
        py::arg("a"), py::kwonly(), py::arg("b"));

[gst]

agree with you py::kwonly looks better.

[skoslowski]

I am also in favor of kwonly(). If there are no objections, I'll amend the original commit.

[wjakob]

This is a nice feature, and the patch looks fantastic. Apologies that it's taken me a while to get to this. Many thanks to @jagerman and @skoslowski.

[henryiii]

Wow, didn't know this went in!!!